R Notebook

Modelo de Regresión

library(tidyverse)
library(readxl)
library(tsibble)
library(feasts)
library(fable)

datos_tbl <- read_xlsx("tiempos_respuesta.xlsx")

New names:
* `Tiempo l攼㹤mite restante` -> `Tiempo l攼㹤mite restante...24`
* `Tiempo l攼㹤mite restante` -> `Tiempo l攼㹤mite restante...27`

#glimpse(datos)

datos_tbl <- datos_tbl %>% 
  mutate(across(contains("Tiempo"), as.numeric)) %>% 
  rename_all(funs(make.names(.))) %>% 
  rename(Orden = N...de.orden,
         Serie = N...de.serie,
         Equipo = N...de.equipo,
         Num.Cliente = N..de.cliente,
         Distribuidor = Grupo.empresarial,
         Nivel = Nivel.de.servicio,
         Recepcion = `Fecha.recepción`,
         Ejecutivo = Usuario.captura,
         IS.asignado = `Técnico.asignado`,
         IS.visita = `Técnico.de.visita`,
         Visitas = N...de.visitas,
         Cierre = Fecha.cierre,
         TST = Tiempo.transcurrido,
         TMO = Tiempo.efectivo.en.sitio,
         TR = Tiempo.de.respuesta,
         TLR = Limite.de.tiempo.de.respuesta,
         TLR.restante = `Tiempo.límite.restante...24`,
         TSP = Tiempo.efectivo,
         TLS = `Limite.tiempo.de.solución.total`,
         TLS.restante = `Tiempo.límite.restante...27`) %>% 
  mutate(Recepcion = as.Date(Recepcion),
         Cierre    = as.Date(Cierre)) %>% 
  filter(Recepcion >= "2018-02-01")

datos_tbl

El modelo de regresión lineal nos permite crear una relación entre 2 variables:

• La variable de previsión \(y\) o dependiente:
• La variable predictora \(x\) o regresora:

Por lo tanto, a partir de esto es como vamos a realizar pronósticos.

datos_tbl %>% ggplot(aes(x = TR, y = TSP)) +
  geom_point() +
  geom_smooth(method = "lm", se = FALSE)

datos_tsbl <- datos_tbl %>% 
  filter(IS.visita != "Consultoria Dual",
         Estatus   == "RESUELTA",
         Categoría == "CORRECTIVO") %>% 
  group_by(Recepcion, Ruta) %>% 
  summarise(across(.cols = c(TR, TSP, TMO, TST), mean),
            .groups = "drop") %>% 
  pivot_longer(
    cols      = starts_with("T"),
    names_to  = "SLA", 
    values_to = "Tiempos"
  ) %>% 
  as_tsibble(
    # index : la variable temporal
    index = Recepcion,
    # key : la(s) variable(s) que identifican a cada serie de tiempo
    key   = c(Ruta, SLA) 
  )

datos_tsbl

datos_week_tsbl <- datos_tsbl %>% 
  group_by_key() %>% 
  index_by(Semana = yearweek(Recepcion)) %>% 
  summarise(Tiempos = mean(Tiempos), .groups = "drop")
datos_week_tsbl

datos_month_tsbl <- datos_tsbl %>% 
  group_by_key() %>% 
  index_by(Mes = yearmonth(Recepcion)) %>% 
  summarise(Tiempos = mean(Tiempos), .groups = "drop")
datos_month_tsbl

p <- datos_tsbl %>% 
  filter(Ruta == "JALISCO") %>% 
  autoplot(Tiempos) + #aes(color = Ruta) +
  facet_wrap(~ SLA, scales = "free_y") +
  theme(legend.position = "none")

plotly::ggplotly(p)

p <- datos_week_tsbl %>% 
  filter(Ruta == "JALISCO") %>% 
  autoplot(Tiempos) + #aes(color = Ruta) +
  facet_wrap(~ SLA, scales = "free_y") +
  theme(legend.position = "none")

plotly::ggplotly(p)

p <- datos_month_tsbl %>% 
  filter(Ruta == "JALISCO") %>% 
  autoplot(Tiempos) + #aes(color = Ruta) +
  facet_wrap(~ SLA, scales = "free_y") +
  theme(legend.position = "none")

plotly::ggplotly(p)

datos_week_tsbl %>% 
  filter(Ruta == "JALISCO") %>% 
  model(ETS(Tiempos),
        ARIMA(Tiempos)) %>% 
  forecast(h = "1 year") %>% 
  autoplot(datos_week_tsbl, level = NULL)+
  facet_wrap(~SLA, scales = "free_y")

Seasonal periods (`period`) of length greather than 24 are not supported by ETS. Seasonality will be ignored.Seasonal periods (`period`) of length greather than 24 are not supported by ETS. Seasonality will be ignored.Seasonal periods (`period`) of length greather than 24 are not supported by ETS. Seasonality will be ignored.Seasonal periods (`period`) of length greather than 24 are not supported by ETS. Seasonality will be ignored.Seasonal periods (`period`) of length greather than 24 are not supported by ETS. Seasonality will be ignored.Seasonal periods (`period`) of length greather than 24 are not supported by ETS. Seasonality will be ignored.Seasonal periods (`period`) of length greather than 24 are not supported by ETS. Seasonality will be ignored.Seasonal periods (`period`) of length greather than 24 are not supported by ETS. Seasonality will be ignored.